Background: Obesity is the major modifiable risk factor for diabetes. This study investigated the incidence of diabetes in relation to multiple anthropometric measures. Methods: Body mass index (BMI), waist circumference (WC), waist-height ratio (WHtR), waist-hip ratio (WHR) and body fat percentage (BF %) were measured among 26 604 subjects (aged 45–73 years) without history of diabetes from the Malmö Diet and Cancer cohort. Results: During 14 years of follow-up, 2935 subjects (1519 men, 1416 women) were diagnosed with diabetes. In men, incidence of diabetes was 24.1 and 4.0 per 1000 person-years comparing the fourth vs. first quartile of WHtR. The multivariate adjusted hazard ratios (HR; fourth vs first quartile) were 6.00 [95% confidence interval (CI): 5.02–7.16) for WHtR, 5.95 (CI: 4.96–7.14) for WC, 5.19 (CI: 4.38–6.15) for BMI, 4.71 (CI: 3.96–5.60) for WHR and 3.21 (CI: 2.75–3.76] for BF%. For women, incidence of diabetes was 15.1 and 1.4 per 1000 person-years for fourth vs first quartile of WHtR (HR: 10.19, CI: 8.10–12.82). HR was 9.16 (CI: 7.40–11.33) for WC, 6.42 (CI: 5.27–7.81) for BMI, 6.75 (CI: 5.52–8.25) for WHR and 5.39 (CI: 4.42–6.57) for BF%. Model discrimination was marginally increased when WC, WHtR or WHR was used in combination with BMI. Conclusion: All measures of obesity were associated with substantially increased incidence of diabetes. Abdominal obesity was associated with higher incidence rates in men than in women, but in terms of relative risks the relationships were stronger in women. The combination of BMI and abdominal obesity measures had stronger association with diabetes than BMI alone.